Homomorphic cameo filter for pattern recognition that is invariant with changes in illumination

Area of Science:

• Optics and photonics
• Image processing
• Pattern recognition

Background:

• Object detection in variable illumination is challenging.
• Traditional methods struggle with illumination inconsistencies.
• Robust target recognition requires illumination-invariant techniques.

Purpose of the Study:

• Introduce a new matched filter method for illumination-invariant object detection.
• Develop a technique that provides consistent correlation peaks under varying lighting.
• Enhance the reliability of target recognition in complex scenes.

Main Methods:

• Logarithmic transformation of the scene to convert multiplicative to additive background.
• Utilizing a composite filter with a target cameo and a discrimination term.
• Implementing a matched filter approach for correlation peak generation.

Main Results:

• Achieved correlation peaks invariant to changes in target illumination.
• Successfully discriminated against false targets using the composite filter.
• Experimental validation demonstrated the method's effectiveness.

Conclusions:

• The proposed matched filter method offers robust object detection under varying illumination.
• Logarithmic transformation and composite filters are effective for illumination invariance.
• This approach significantly improves target recognition performance in challenging visual environments.